JPS60500472A - Method and apparatus for processing speech signals - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed description of the invention] Method and apparatus for processing speech signals TECHNICAL FIELD The present invention relates generally to the processing of speech signals, and more particularly to the information content of speech signals. Method and apparatus for amplifying speech signals for hearing-impaired persons without adversely affecting Pertains to.

Hearing-impaired people generally have significantly limited access to Quinamine cleansing, i.e. Speech intensity levels that are barely audible (e.g., speech recognition limits) Speech intensity level that is most easily audible (maximum) appropriate level (MCL), and a speech intensity level that is too loud to be heard (uncomfortable). It has been found that the difference between the volume level and the volume level [JCL] is usually small. Therefore, In general, a wide range of speech intensity levels can affect hearing. Suitable for each individual with There is a strong demand for lowering the range to a limited range. However, peak chestnut Up and instant warning? Also, in the public input 11C) device using Emphasizes the strong low jh no kaku l), and makes the high-pitched Chou Shengqi faint with an antagonistic distorted force cry. Ruru. American Speech, Rock Hill, Marilant, April 1979 Written by Mr. BrXLida et al. γ1. llearning aids-A Review of Pa5t 1lesearCh onLinear AmplificaLio n, 8mpliLude Compression, and Freqllen C) Extensive research results are presented in the literature entitled 'Lowering)'. . , :'s literature provides guidance on the concept of compression as well as a wide range of rigorous information on pressure♀j6 It has been reassigned to wealth. C) In the literature, the effect of J-Una compression shown here is If this cannot be obtained, there is no guarantee that the amplitude compression itself will not be effective. 811.1 The examiner side explained the concept of pressure pipe 1 and its complexity δA+1'J; It is suggested that this is because it is difficult to fully understand in practical terms.

The nine patterns of speech sounds are systematically divided into three main time domain components G as follows. I know that it can be analyzed. (1) Display the spectral carton of each short sound segment. The slight temporal pattern that I am experiencing, +21, the change in the second minute temporal pattern. rough temporal patterns representing the duration of various sound segments and (z31 time-varying amplitude pattern. 5-10 ms steps) Subtle temporal cues originating from spectral segments identify consonant pronunciation positions often give the listener enough information to Similarly, a rough temporary pattern should provide sufficient information on how to pronounce the sounds, especially the first part of fricatives, affricates, and plosives. Give often. Amplitude pattern that changes over time, instantaneous signal envelope line Is it a result of the speech formation process G? The most redundant emotion (I can also be conveyed) that has already been conveyed through temporary patterns. . Mr. Robinson and the former Mr. Longton, 1973 ustical 5atIeLy　o(Am,: of the conversation that took place before rica) Among them, in the compression amplification of the following deviation, undesired distortion may sometimes intervene due to the use of a short time constant7. On the other hand, he acknowledged that if the time constant is long, it becomes too slow.

Continuously calculate the average power of the speech waveform at intervals of several tens of milliseconds, Using A method has been proposed in which the oscillating pressure 116 signal is delayed by two half the length of the averaging interval. Ko If you obtain the computer simulation lane in advance, the speed will be Chi no Shimen Sano/Konobu 1: It has been suggested that Pigzen can be done with I Seno. However, Mr. Rahinson and others! : Mr. J’1lun1.1nBLθn explained this process. A: Research on research (I haven't done more than 3) stomach.

The purpose of the present invention is to easily dry the speech. Improvement 6 and 7: To provide a speech processing method.

Still another object S4 of the present invention, fine silk nine temporary turn and rough speech Temporal patterns should have negative effects, but t6<, amplitude patterns that change over time That is, adjust the speech pattern to minimize changes in the envelope! + equipment It is to provide.

The present invention provides an average value or an effective value (R to Is) amplitude pattern that changes over time. and use stiffness to normalize the amplitude pattern that changes over time of the speech. Ideally, the listener's optimal level is between the speech recognition limit and the uncomfortable volume level. A speech processing method and apparatus for forming a compressed speech pattern located at the bell. It flashes everywhere. Spectrum shaping is done to make it high )stomach). Ki Jōmei; 2, ri-chirente', l! , '4: (,: Totally 1 min... ,・1? (Using the device 2-Kinoshii LJ tooth part 1'1', 10 Honshi Dan) Speech pattern (Kalu Ward) FB → Rasai 9, microphone 1 cohon to C: I Subina Putter> I; 1 extra] 41 first 5, and appropriate shaping fleet, Inino 1 evening processing 7 Well, Takashu's boyfriend is going to be in a bad mood.

Is this sub-shaping or the effective signal amplitude III (RMS)? J! F5t time period 'ζ泶1sachi1, this effective (lit the reciprocal of J, space A V-shaped beam is modulated and a normalized amplitude is formed. important things Before amplitude compression, the shaped signal has an effective value of /V! Time delay included in 1 hiki The delay will be sufficient to compensate for the delay. The signal generated by this is pressure 2. -15, by blowing, spectral shaping produces minute temporal butterflies and 88 2'1° so that the temporary pattern is retained.'' - No. 13 or desired The purpose of the present invention and its objective skin characteristics 1″j are as follows: 〇, J who explains the plan of F after V, and i of the claim, Bashiriki, etc. will become clear.

11th Mele, one of the unexploded 1st sons It is a 0″J function block diagram showing the nadokoro■v device.

2nd] Zui: 7-(There is), which shows the first part of the speech based on the present invention.

No. 31, Multi-channel implementation 1 of the speech processing device according to the present invention 1l-2 (Mr. ``1 Hifu l-I Nokre 1.

FIGS. 4A and 4[3] show the function of a tape recording device according to another embodiment of the present invention. This is Ronocre 1.

Without explaining the accompanying drawings, Fig. 1 shows the present invention constructed using general heavy enclosure parts. FIG. 2 is a functional block diagram of a sato-nayanne spinner processing device according to an embodiment of the present invention. child Implementation i + sc Here, a microphone 10 with wide frequency response characteristics (e.g. For example, Knoles F and A that control the response characteristics of 100H2 to 10KH2 1934-like electric microphone 2.l.microphone) ↓ picks up the audio signal. -(,10(lN).

Not Llol, III, T1 River S, l with a gain of 26 dB; 1 Place magnifier 12・＼Electricity (Jx 5 bows. Increase + j + sa 1st Jinshi, then (, high Frequency bow I! Glue d 14 each (1 sho 1: Eha 'l' l Ob 4 rin (t increase 1↑J device 9-・Sui, γ1, this old and old, the sum of l00KII□Muishi2KH, The woman i for I is 6d1: / the gain of off-turf is Ij, and 2 k l-1 , less than '-' - frequency G is set to ↑ to give a flat response. For example, a Lan-O receiver? An auxiliary input is provided at position 164 so that these signals can be supplied to the high-level $ emphasis circuit 14. I'm being kicked.

The signal from the high/dark wave strong li+ circuit 14 is the RMS detector 16-/send port I.

and a delay circuit 18 which provides a delay equal to the time constant of the RMS detector 16. It will also be sent to - Considering the example, RM S + = Output RX or Atero 4 series △l-) 536 A composition 18Jf, and soon le i11 road 18 or reticon (1leticon) SA D 4096/’ (nonon relay type device ζ+ This device is operated by Denotal Chronok 20 of 8oKH2. .

The delayed rl signal from the delay circuit 18 is sent to the divider circuit 20 (for example, the analog Vice? t: (Analog li: vices) 0) A D 535 high-precision μs divider) ．． 1 M　S swing +11 or Fr Duke and U7 defense, 9 units/\send. (talk, cut Seriously, the output from the device 20 is delayed and the signal is put under pressure ♀tW.Is this it? The signal is sent to the receiver 22 (Knowley, El 1925).

Figure 2: Input level of signal from microphone 10 dBS1〕1- , plot the compressed output level d133 I) L of the signal sent to the receiver 22 This is what I did. If the input level is lower than approximately 45dB, the output level will be attenuated. be done. When the input level is 45d 13, the output level is compressed to about 1 00dBsPL □ This is L: Mc +-, the level is uniformly dark blue 71 .

dl < -4: 1') Input to A, 71.4 to -C is pressure I LZj"　I　O:　l　j1', 2; 9!: Eve') 5 or more〇Please keep it coming .

No. 3 "Figure 86, the first part of the multiplication, one part of the skin surface, and one part of the skin. Fill 1 + 10 years old Yagi 4 is another implementation of Akira (lid Koyoru multi-channel) 2 shows the Leu-1L pressure τ10 device. In this 'A example, the microphone 30 32, mid-range (400-1 6001" 2) Solder 1' router 34 and high frequency (+600-6400H7) filter 3 Sent to 6. Each fill [signal shift from evening 11'l, 1. 'fi Inako Mτ 38. 40 and 42 ＼ feeding and discharging). Each pressure path is shown in Figure 1. It is equipped with a delay circuit, an RMS detection circuit, and an initial division circuit. Each Chi No Fu Since the color of the narrowband frequency is gold, the high position in Fig. teshi)ru. Pressure sheath 1 year old 1. i) 1. Yu Z, Oho Leopard Increase [Slow Device 714・\Sendshi, 27. Illusion (1α sa 2 yō, kikugin 1100″-i or uke (divination device, 16, ＼Miran').

4th A 1p, i and 413th jgl'=', :, Chi77' l-yo Not used with ku; ノH (-useful j6 rice invention G) to show another husband bΦ1 or Ij −Zl' l:ltsuk1nu1, in this case, the compressed signal and the detected Bo1. TA1ZE~I S (II "1 both or time 1SJc: 21111'j recorded on the data recorder. In the case of Figure 4A, G is the microphone '50 or signal from another audio source, or a The signal is then sent to a width compression device 52, which is a multi-channel device. Pressure voice message The number is recorded in the analog recorder 2 of the cheap recorder 54, and the detected R The MSI 16 is recorded in the FM channel of the recorder 54. Then record compressed audio signal and recorded l? The M S value is sent to the multiplier 56, From this, the original audio signal and the original dynamic range are formed. By kneading The resulting decompressed signal passes through a frequency strength 111 cancellation circuit 58 to a receiver 59. Sent.

Fig. 413 shows a 4-noshiburi 7' similar to the analog l-v reversing device in Fig. 4A. (indicates a tape recording device). This hell is in one line ↓, t, microphone l:!　10＞60? Their faith: No. 1! Same as in the case of A i': l　2・Tsuji et al. 3', , Greed: I, sweat, Hakusai 1. ’G’ I”’lR゛ ・J and r-\+<xi s value cuff -J-1:l 1' 5-Sotal conversion circuit According to 63 and 65,' The digital signal is recorded on the tape recording device i'Z:64L.

This recording was borrowed from Tenotaru-7sljkukonharta57yp'2'1 １　D　AC　C:　661 will be used again. 1 −) The decompressed signal from △066 is processed by the frequency ζ and jqj key removal circuit UfH6 8LIJ is the emphasis removed, and then sent to the receiver 69 by Jx.

Cheap recorders typically have a dynamic range of 1 (R constant 4j), so These embodiments of the invention are in 'l11'. Therefore, the pressure ei? Sa2ma Tabata-1 is also complaining, 'Q'! ? xi S jl? + to the recorder G All the recorded signals 7) are 41 mo 9 equipment 50 Lupi 66 Oko right + j T + I ~ 11 戊, 2, etc. can be done.

1111 ko J ho \f 2 preferred P: Is falling on the side l? M:) For one person did.

However, the time 17; , an average of 111 supports M S 41'l (including +ffl) should be used. Mote@ru. Here, we use Zuro 1ze M S (11+ is the appropriate Kin 1 Kei (lI! For example, to obtain a compressed signal, a divider is used. The force used was detected, I'? M　S　I1ri or the logarithmic scale of the mean You can also use it to obtain Pressure White Sen.

The present invention can be applied, for example, to hearing aids and audio recording 1a media (such as recording L), sampling type General audio communication equipment, including digital storage, broadcast systems, public address systems, and telephones. It can be used for a wide range of purposes such as:

The present invention is useful for communicating in noisy environments or for use in the field. This is particularly useful when communicating via a different communication link.

A certain fruit hi! ! example? ,'Explaining the present invention-Snib 1) Two-burning invention/ One! There is something to explain the uninventive nature of the chrysanthemum.

The true merits of the invention as defined in the claims! Don't deviate from Ukuno゛・ra Numerous historical changes 11. A person skilled in the art would be able to obtain the shortness γ1 (or C1).

20 40 60 80 100 Input of 14 circuits “Ch: dB 5PLFIG, -2 FIG, -4A FIG, -4B Fourth amendment (method) %formula% 2. Title of the invention: Method and device for processing spinner signals 3. Person performing correction Relationship to the incident: Applicant Name: Henri Noxon, 1 other person besides Larry Keith 5. Date of amendment order: December 25, 1980 Engraving of the translation of the drawings (changed in content) none).

international search report

Claims (1)

[Claims] 1. In a device that processes audio signals including speech signals, a selected time period means for obtaining a measure of the amplitude of the spinner signal during and the audio during the selected time period. and means for compressing a signal based on the amplitude scale. Place. 2. The means for obtaining a measure of the amplitude comprises a root mean square (RMS) detector. A device according to scope 1. 3 Juki effective value selection u3 instrument or the above selected time rg period, the above sound μ month, 1 possible [1 effective 1+il! 141 signal during the above selected time period. 3. Apparatus according to claim 2, further comprising means for delaying by . 4.--While the above measure of the amplitude of the audio signal is obtained during the selected time period, the sound Delay means for delaying the voice signal by the selected time period (III) The device according to Box No. 1 Jβ. 5, 'r'N, 1'.Receives and responds to the 5 lever phase: My device that emits an electrical signal. [1. This microbond means Z,X connection is used to amplify the electrical signal mentioned above. ; 1 wave number emphasis IIT means and this high IN 1 wave number emphasis means - (, increase ] ↑J-sa? '1.1 Receive this signal and perform this amplification for the selected time period. vibration 11 detection means for obtaining the effective (RMS) amplitude of the signal and high frequency emphasis means. A connected ear, a spectrally shaped force, receives an electrical signal, and transmits this signal. delay means for delaying the signal by the selected time period; ” - Record I2i! connected to the spreading means and the swing width detecting means, and increases the upper gF [1] S>”L Delayed box: Compress the electrical signal by compressing it to the above effective average 11 A voice processing device characterized by incorporating a signal compression means. 6. The above-mentioned high wave number 1°, 18] Means B, H1 above 1°,) at least 2 to 1-1° is increased by 6tl13/octave Claim No. 51ji The device described in. 7. According to claim 5G, the delay means is a relay type device. The device described. 8. The compression means converts the delayed non-linear increase rl+i3 into an effective value An apparatus according to claim 5, comprising a divider for dividing by . 9 Prepositional addition of the above microphone means to the above Takagokami, 1 means 6 connection U +4] The device according to claim 5, further comprising #iA means. 10. When selected in the method of compressing the amplitude of the signal within the audio signal processing device Obtain a measure of the amplitude r[J of the signal over the interperiod period, and The above-mentioned signal corresponding to the above-selected time period is I took 16 balls and left one temple and Zuroka l. 111-ki 41: Mountain Shake (Color ° Kaisho Ima 1 ``I'll keep it straight!'' 1%f, method of 2nd record 11M. 12 The compression stage described above, when selected, causes f! j:　,lRj　! ! Jl 6 Nigen 1) The first range of a range includes dividing the message by the person's effect + t1. The method described in Section 1. 13. Before obtaining the effective value, further step (I The method according to claim 10. 14. Receive the audio signal and pass it through a plurality of fixed frequency bands and filters. A bandpass filter with a factor of 1 and this band i[i! each connected to a passing filter Equipped with a powerful swinging width compression means! +iM, and each width compression means is 1 when it is ashed. Select the means and the sound signal to obtain the amplitude of the audio signal [11 degrees] during the 111th cycle. Sawata time 11th period 1st fall! ! Means to extend and during the above time period! ! extended 10 means for compressing the audio signal based on the amplitude measure, and further , the above-mentioned plurality of swinging pressures of 4t? 1 means, and receives the audio signal with pressure i+1. An audio signal processing device characterized by comprising an adding means for receiving and adding. 15. The above-mentioned means for obtaining the measure of vibration 111 is 'effective (1 force (121ψS)) detector. 15. A device according to claim 14, comprising: I6, record the audio signal and obtain the original dynamic range from this recorded signal In the method, obtain a measure of the amplitude of the audio signal over the selected time period; Compress the audio signal corresponding to the period based on the amplitude scale, records compressed audio signals, and recording a measure of the amplitude of the audio signal. 17 The above amplitude (7) scale is the effective flu 'C;'84 jil im, "il'i l 6IQ: Koji @Q 7' left. 18 Compressed audio recorded No. 13 and Record C: 21mu = Ichiko Shinyoku's Shaku 1 boat and Search for ``2. Recorded voice signal amplitude [1] Using scale'' - Recorded eyebrow recorded neck Claim 16 includes a method for decompressing a signal. 19. In the recording step described in 1, the compressed analog signal and amplitude scale 1!1 are digitally recorded. The decompression step described above includes converting the recorded digital 19. The method of claim 18, comprising converting the digital signal into an analog signal. 20 The above recording step converts the compressed analog signal and amplitude measure into digital form. 17. A method as claimed in claim 16, comprising converting into a state.